Analysis And Design Of Information Flow Advertising Recommendation System Based On User Personalized Data

With the vigorous development of social network media and big data mining technology,the information flow advertisement recommendation system based on user personalized data has gradually become a key research project of many companies.Information flow advertising is a field involving advertisers,advertising platforms and Internet users.Due to the different goals of the three,advertising platforms closely integrate advertisers and Internet users.Advertisers pursue a return on investment ratio,and Internet users pursue the experience and enjoyment,the advertising platform is pursuing profit and income,and the information flow recommendation system is a concrete commercial product.This system makes the information flow advertising from the original single push to the current personalized recommendation,the advertising effect is multifaceted,and the efficiency of advertising promotion has been greatly improved than before.Nowadays,there are many researches on the information flow advertising recommendation system in the Internet business field,but the related research is very scarce and outdated,and the recommendation accuracy and system throughput are poor.In order to solve the above problems,this article mainly carried out three researches on the technology of information flow advertising recommendation system:1.The information flow advertisement recommendation system involves the operation of matching a large number of advertisements with users.In order to solve the performance problem of the recommendation system,an information flow recommendation design based on multi-channel parallel recall and unified sorting distribution is proposed,and the design is discussed in terms of high expansion,low latency,and strong stability.2.Recommended system accuracy.The core evaluation index of the information flow advertising recommendation system is the accuracy of recommendation,that is,the right content is provided to the right users at the right place and at the right time.In order to improve the accuracy of recommendation,this article introduces some traffic strategies for accurate advertising in daily life,and introduces the practical application and effect evaluation of Light GBM,XGBoost,and Wide&Deep in this system.3.Advertising system revenue.In order to maximize the revenue of the advertising recommendation system,this article proposes an advertising bidding strategy based on model weights,product weights and advertising bidding value,indicating that this strategy guarantees the balance of user experience and advertising revenue to a certain extent.This thesis designs and implements an advertising recommendation system based on user personalized data flow according to the above mentioned related technologies and theories,which can efficiently and accurately recommend the content that users are interested in.Experiments show that the relevant algorithms and engineering strategies we use are reliable in improving the recommendation accuracy and system explosion of the recommendation system.